CAPITAL 1 POLÍTICA CONTABLE

This study attempts to elucidate the factors that contribute to facility entry and exit. Many of the same explanatory variables (e.g., competition, procedure demand) contribute in different ways to each of these competitive actions so each is discussed within its own heading. However, these explanatory variables must also be measured at different levels of observation for the purposes of testing our models of facility exit (facility level) and entry (county-specialty level). Each sub-heading establishes which model is being described and how we measured these variables for the model in question. Tables presented along with our description of the empirical models should help by simplifying the variables names and descriptions as well as outlining the models they appear in.

A. Competition 1. Facility Exit

Several of our hypotheses (A4, B4, A5, B5) relate the likelihood of facility entry and exit to the level of competition within local health care markets. This piece of the analysis requires that we measure the density of surgery centers and general hospitals with overlapping and non-overlapping niches. In order to do so, we borrowed heavily from Baum and Singh, who dealt with similar issues in their study of Toronto day care centers (Baum & Singh, 1994a, 1994b). Their study constructed a matrix of niche overlap weights that depended on the extent to which specified age ranges tended to

overlap with one another. A similar formula guided the construction of the weights employed here: i ij ij ij s s s w + = j ji ji ji s s s w + =

where sij and sji represent the number of procedures they perform within the same

surgical specialties while si represents the number of procedures performed by firm i in

specialties that firm j does not participate in and vice versa.10 The weights calculate the extent to which their procedure volumes overlap.

Baum and Singh (1994) measured the fundamental niches of child day centers by constructing these weights based on the ages of children that day care centers were eligible to enroll and so a different weight was applied to each class of day care centers. Their implicit assumption was that each age group constitutes an equivalent proportion of each day care center’s business. An identical approach here would have involved

constructing weights based simply on the surgical specialties that each facility

participates in. However, we know that there may be significant differences between these organizations’ realized niches and their intended niches since each specialty that a facility participates in does not contribute an equal amount to its total procedure volume. Although we would have ideally based these weights upon revenues, we were able to construct far more detailed weights based upon each facility’s procedure volume in each specialty with the availability of more detailed patient data.

From our niche overlap weights, we calculated the overlap and non-overlap density for each outpatient facility from the following formulas:

∑

= = n j ji i w density overlap 1

∑

= − = n j ji i n w density nonoverlap 1

where wji represents the weights that were described previously and n represents the

number of outpatient facilities in the health service area occupied by firm i. To calculate overlap and non-overlap density for firm i, we counted up the number of facilities in its health service area at time t and then assigned them all fractional weights based on the extent to which their procedure volume overlaps with that of firm i at time t. Overlap density is the sum of all these weights and it is intended to represent the aggregate proportion of all facilities in firm i’s health service area with overlapping niches. Non- overlap density was calculated by subtracting the sum of these weights from the total number of outpatient facilities in firm i’s health service area and it represents the

aggregate proportion of all outpatient facilities in a market with non-overlapping niches. As we have previously described, we broke down our measures of competitive intensity into their constituent components. We began by measuring these overlap and non-overlap density at the highest possible level and then calculated each measure separately for ASCs and hospitals to account for differences in facility type. We subsequently divided overlap and non-overlap density into their local and diffuse components in order to observe the effect of geographic location. Finally, we utilized

both of these distinctions – facility type and geographic location – simultaneously in order to observe their combined effect.

2. Facility Entry

In contrast with our facility exit models, the unit of observation for our facility entry models is the county-specialty. In the case of facility entry, the facility does not yet exist and so we modeled facility entry as a function of each specialty’s characteristics within each county. Although competition has opposite effects upon facility entry and exit, our method of calculating competition at the county-specialty level was derived closely from our method of representing competition at the facility level. As with our facility-level niche overlap weights, we divided each facility up by the proportion of its procedure volume that it devoted to each surgical specialty. Following the example of Baum and Singh (1994), we constructed niche overlap weights from the proportion of each facility i’s procedure volume that was devoted to specialty j within county k:

(1.1)

∑

= = ₁₅ 1 j ijk ijk ijk s s w

where sj represents the number of procedures that facility i performs within specialty j.11

Given these weights, the overlap and non-overlap density within each county-specialty was calculated from the following formulas:

(1.2)

∑

= = nk i ijk jk w density overlap 1

∑

= − = − nk i ijk k jk n w density overlap non 1

where wij represents the weights that were described previously and n represents the

number of outpatient facilities within county k. Adding up our niche overlap weights by county produced the niche overlap density, which represents the fractional number of facilities that served a specialty within each county. Adding up niche overlap density across all surgical specialties yielded the total number of outpatient facilities within each county. By subtracting niche overlap density from the total number of outpatient

facilities in each county, we were able to calculate niche non-overlap density.

These density measures were computed differently for our facility entry and exit models because these models utilize different units of observation. The unit of

observation for our exit models was the facility-quarter and so we calculated niche overlap and non-overlap for all facilities competing with the focal facility within a given quarter. The unit of observation for our facility entry models was the county-specialty and so we calculated niche overlap and non-overlap for all facilities operating in the focal county-specialty within a given quarter. Admittedly, these measures are somewhat sensitive to market size, so we experimented with alternative measures of competition and calculated a Herfindahl-Hirschman Index (HHI) by facility and county-specialty. We inserted HHI into our facility entry and exit models instead of our measures of overlap and non-overlap density. However, we found that HHI produced no consistently

significant effects for either ASCs or hospitals.12 We even included these HHI measures along with overlap and non-overlap density order to observe whether our observed effects

12 _{We found that HHI has no consistent effect on ASC entry/exit – the coefficients switch signs – and has}

disappeared but found that they were robust to the inclusion of these additional independent variables.

We included county-specialty level measures of niche overlap and non-overlap density within our facility entry models. As with the competition variables that we included in our models of facility exit, we gradually broke these variables down into smaller components. Initially, we calculated these figures at the level of the HSA for our first model. Then, we broke them down by facility type in the second model. In the third model, we assumed that the county-level variables represented local competition and that diffuse competition was represented by all other counties within the HSA. Finally, we introduced both distinctions – facility type and geographic location – simultaneously in the fourth model to examine their combined effect.

B. Procedure Demand 1. Facility Exit

Having established our measures of niche overlap and non-overlap density, we included a number of additional variables in our models in order to control for a number of other factors that might affect a facility’s likelihood of exit. First and foremost was the size of demand within the surgical specialties that a facility serves, which represents the environmental carrying capacity. In order to measure market demand, we generated quarterly procedure counts by patient county code instead of by facility code.13 We then aggregated these procedure counts across each specialty that a facility participates in

13 _{We recoded out of state residents with the county for the facility in which they were treated, which was}

based upon our definition of specialty participation. This procedure generated two separate sets of procedure counts; we included outpatient procedure counts within our exit models for ASCs and hospitals while we included the inpatient counts for hospitals only. These procedure counts were deemed to be a reasonable proxy for the level of demand within the markets that a facility serves. As with our measures of competition, each of these demand statistics was broken down by facility type (ASC vs. hospital) and geographic location (local vs. diffuse).

2. Facility Entry

Levels of procedure demand are likely to affect the likelihood of facility entry and so we included in our entry models a variable representing the size of demand within each county-specialty, which represents the environmental carrying capacity. Once again, we generated quarterly procedure counts at the county-specialty level from patient county codes, and these procedure counts were deemed to be a reasonable proxy for the level of demand within each market. We generated these quarterly procedure counts from both the inpatient and outpatient datasets, which produced two separate sets of procedure counts; we included outpatient procedure counts within our entry models for ASCs and hospitals while we included the inpatient counts for hospitals only. As with our measures of competition, each of these demand statistics was broken down by facility type (ASC vs. hospital) and geographic location (local vs. diffuse).

C. Physician Statistics 1. Facility Exit

The setting for this study is unique in that outpatient facilities not only compete for procedures but also for physician referrals. It is typically surgeons that refer patients

to a given outpatient facility and patients rarely disagree with their choice of facility (Kouri et al., 2002; Lynk & Longley, 2002). These relationships that exist between surgery centers and their physicians may affect exit probabilities such that a facility’s likelihood of exit depends on the extent to which the facility depends upon its surgeons and vice versa. To that end, we included some measures that control for these

relationships.

Conveniently, there is an ID code representing the operating physician in each record of the outpatient dataset. These physician IDs were cross-tabulated with facility IDs in order to determine where these surgeons are performing their procedures. By generating these quarterly procedure counts, we were able to ascertain how facilities divide procedure volume across surgeons and how surgeons divide procedure volume among facilities, which allowed us to produce a variety of control variables that depict the multi-dimensional relationships that exist between facilities and their physicians. We also experimented with weighting these measures by each surgeon’s contribution to facility procedure volume in order to give more weight to high volume surgeons. However, this approach had no discernable impact on our results and so we included the unweighted measures in our models.

To begin, we produced a simple count of the number of surgeons operating within a facility in any given quarter. We also included a control that represents the average number of facilities that the facility’s surgeons are utilizing for their outpatient

procedures. Taking into account procedure volume to represent facility and physician dependence on one another, we also produced a measure of physician-facility inter-

dependence (PFI) (Feldman & Wholey, 1999; Marsh & Feinstein, 1997; Wholey & Burns, 2000). It was calculated according to the following formula:

i facility to admitted j physician for surgeries of Percent j physician from coming i facility in surgeries of Percent FI P = j physician by performed surgeries of number Total i facility in j physician by performed surgeries of Number i facility in performed surgeries of number Total i facility in j physician by performed surgeries of Number FI P =

Higher values of the PFI indicate that the physician has relatively more leverage over the facility while lower values indicate that the facility has more leverage. We included within our model the average PFI across all physicians operating within a facility.

2. Facility Entry

The availability of surgeons is likely to contribute significantly to facility entry and so we experimented with the inclusion of a measure representing the supply of practicing surgeons within our entry models. Our method of calculating the supply of physicians within each county-specialty proceeded in same manner that we employed when calculating facility niche overlap density at the county-specialty level. We began by dividing up each surgeon’s total procedure volume by county and surgical specialty. Using the dataset that we cross-tabulated by facility and physician IDs, we attributed surgeons to the county in which their outpatient facility resided and calculated separate overlap weights for each county in which a surgeon practiced.14 We then added these niche overlap weights in order to produce the fractional number of surgeons that operated within each county-specialty.

We considered including these measures of physician supply in our entry models and we also considered weighting them by procedure volume in order to distinguish between high- and low-volume surgeons. However, we ultimately determined that both the weighted and unweighted measures were highly collinear with our measures of procedure demand and so we excluded them entirely from our entry models. As an alternative, we considered modeling facility entry and exit as a function of surgeon entry and exit but were unable to do so. The quarterly number of procedures at the facility- physician level was so low that our method of calculating physician entry and exit produced a large number of false-positives and false-negatives; without physician licensure data, was no way of validating these entries and exits like we did with outpatient facilities. We opted not to pursue this approach rather than including such erroneous measures of physician supply.

D. Control Variables 1. Facility Exit

We also included a number of facility-level control variables in our model that emerged from our panel dataset as well as the Florida licensure data. For example, we previously mentioned the “liability of newness,” which suggests that relatively

inexperienced firms are more likely to exit than firms that have more experience (Freeman, Carroll, & Hannan, 1983). To control for this effect, we added a covariate representing firm age to our model. Firm age was calculated from the entry dates

provided by the Florida licensure dataset, which we validated with the dates produced by our outpatient procedure volume dataset.

Although surgery centers and hospitals occupy different macro-niches, there are relative specialists and generalists among each type of organizational form. For example, some multi-specialty ASCs perform every imaginable surgical procedure while there are specialty hospitals that participate in only one surgical specialty. To that end, we

included a variable to control for the possibility that specialists or generalists among either organizational form may have a higher likelihood of exit. We calculated the number of surgical specialties in which each facility participates – based upon our definition of specialty participation – and we included this variable in our model to control for this effect. Alternatively, we experimented with dividing our sample of surgery centers and hospitals in half based upon the median number of specialties in which they participated and including dummy variables representing the specialists and generalists among each type of facility. However, we ultimately decided to include our continuous measure of organizational specialization because it offered much more detail than these dummy variables and maintained a closer correlation with our measures of competition and market demand that were based upon the specialties in which a facility participates.

We controlled for the fact that some procedures are more generously reimbursed – and are therefore more profitable – than others and that this characteristic is likely to affect a facility’s likelihood of exit, depending on the specialties that the facility serves. Although we would have ideally measured reimbursement-cost ratios directly, we were unable to generate data representing either reimbursement rates or average facility costs by procedure type. Instead, we included variables representing payer type under the

presumption that certain types of payers reimburse providers more generously than others. The Florida outpatient dataset also includes a field that describes the payer for each procedure in the dataset. Using the payer field, we cross-tabulated facility

procedure counts by payer in the same way that we did to generate physician statistics. These procedure counts can be divided by a facility’s total procedure volume in each quarter in order to calculate the proportion of a facility’s procedure volume that is reimbursed by each type of payer. Although there are a total of 16 different payer codes in the Florida outpatient dataset, we collapsed these codes into three categories: public, private, and other.15 _{Variables representing the proportion of a facility’s business that is}

reimbursed by public and private payers were included in our models while other payers were omitted from our model as a comparison group.

We also included several other control variables that emerged from facility licensure data provided by the state of Florida. For example, the state provided data on the size of the facilities in our sample. For surgery centers, size was represented by the number of operating rooms; for hospitals, it was represented by the total number of beds.

In document Elaboración e implementación de manual de políticas contables (NIIF) y sistema organizacional con auditoria fiscal para la empresa Metroin S A S (página 155-160)